HighSpeed is the fast lane for Kimi K2.7 Code, Moonshot AI's coding model. It is the same model giving the same answers, only written out much faster, for twice the price. This page explains how fast it is, what it costs on the API and in Kimi Code, and when the extra cost is worth paying.
An illustration. The bar lengths are not measured timings.
Kimi K2.7 Code, served at about 180 tokens a second.
Double on the API. About three times the credits in Kimi Code.
When you are waiting at the keyboard. Not for background jobs.
When a coding model works, you watch it write: plans, code, explanations. With a slow model, a long answer can keep you waiting. HighSpeed runs Kimi K2.7 Code on faster serving so the words arrive sooner.
Moonshot describes it as the same model. Same knowledge, same skills, same answers.
About 180 tokens a second, rising to about 260 in short conversations.
Twice the standard API price, and about three times the credits in Kimi Code.
| Standard | HighSpeed | |
|---|---|---|
| Speed | Normal | About 180 tokens a second, up to 260 in short conversations |
| Answer quality | The same model, so the same answers | |
| API input | $0.95 | $1.90 |
| API cached input | $0.19 | $0.38 |
| API output | $4.00 | $8.00 |
| Batch price | Yes, 60% of standard | Not listed |
| In Kimi Code | Paid plans | Allegretto or higher, about 3× the credits |
| Reads at once | 262,144 tokens | |
On a small screen, swipe the table sideways to see every column.
API prices per million tokens, before tax, from Moonshot's price page. Kimi Code details from Kimi's help centre. Moonshot does not publish a speed figure for the standard version.
A token is a small piece of text, roughly three or four characters. At 180 tokens a second:
200 tokens, a paragraph or two, appears in about 1 second.
800 tokens of code appears in about 4 to 5 seconds.
5,000 tokens, including thinking, takes under 30 seconds.
Simple arithmetic at 180 tokens a second, ignoring the wait before writing starts and any time spent running tools. Real speed varies. Kimi K2.7 Code always thinks first, and thinking counts towards these totals.
Change the model name in your request. Nothing else changes.
kimi-k2.7-code-highspeedapi.moonshot.ai/v1Pick it in the Kimi Code tool, if your plan includes it.
/usageUse a Kimi Code key in tools such as Claude Code or Roo Code.
kimi-for-coding-highspeedapi.kimi.com/coding/v1Full setup: K2.7 Code API · Kimi Code · Kimi API setup
How much waiting would HighSpeed save you, and what would it cost? Enter a typical day on the API.
Moonshot does not publish a speed for the standard version, so enter what you see. The default of 60 is only a placeholder. The calculator assumes HighSpeed writes at 180 tokens a second and counts output only. Input costs double too.